Several methods have been proposed for guiding pedestrians
by generating guide sentences from navigation routes
[10], [11]. These methods use landmarks to explain the
next point to which the users should go. However, in these
cases specifying the landmarks is difficult because they use
only names as guide annotations without any exterior form
of identification. Hattori [12] proposed a utility that uses future work, we will consider how to select the best guide
annotations from those submitted by users.